A Distributed Routing and Time-slot Assignment Algorithm for Cognitive Radio Ad Hoc Networks with Primary-User Protection

Transcription

1 2012 7th International ICST Conference on Communications and Networking in China (CHINACOM) A Distributed Routing and Time-slot Assignment Algorithm for Cognitive Radio Ad Hoc Networks with Primary-User Protection Hao Chen, Pinyi Ren, Li Sun and Qinghe Du School of Electronic and Information Engineering Xi'an Jiaotong University, Xi'an, China js.sq.chenhao@stu.xjtu.edu.cn.{pyren.lisun.duqinghe}@mail.xjtu.edu.cn Abstract-Cognitive Radio (CR) technology enables Secondary Users (SUs) to transparently utilize the licensed spectrum bands while causing limited interference to Primary Users (PUs). In this paper, a distributed routing and time-slot assignment (DRTSA) algorithm is proposed to address the PU protection issue. Specifically, we first design a proper routing metric for distributed path selection and then we propose a time-slot assignment algorithm to improve SUs' capacity and reduce the interference to PU as well. Theoretical analysis shows that the complexity of DRTSA is no more than O(NTlogN), where N and T denote the number of nodes and the time-slot period in the network respectively. Simulation results show that, under various network scenarios, the performance of the proposed distributed algorithm can approach that of the centralized alternatives. Index Terms-Cognitive Radio; Distributed Routing ; Timeslot Assignment; Primary-user Protection. I. INTRODUCTION To make sufficient use of the spectrum resources, the notion of Cognitive Radio (CR) was proposed by Dr. Joseph Mitola in 1999[1]. Since then cognitive radio has been widely considerated as a revolutionary technology for next-generation wireless communication systems [2]. Due to the high fluctuation in the available spectrum, cognitive radio imposes unique challenges to current communication systems. Specifically, in Cognitive Radio Ad Hoc Networks (CRAHNs), the distributed multihop architecture and the dynamic network topology make the network design complicated and have attracted increasing research attentions [3]. In CRAHNs, Primary Users (PUs) have high priority in the utilization of the spectrum, while Secondary Users (SUs) are allowed to sense and explore PUs' spectrum and identify underutilized spectrum blocks for data transmission. To optimize the performance of CRAHNs, researchers have conducted many studies on spectrum sensing and MAC protocol design. And a lot of progress has been made on these two fields [4] [5] [6]. However, MAC protocol which gives optimal solutions in the single hop configuration may become highly inefficient in a multi-hop CRAHNs [7]. The authors of [8] extended the routing solution in Multi-Channel Multi-Radio Networks (MCMRNs) to CRAHNs and proposed a layered graph framework to address channel assignment and routing jointly. Hou et al. in [9][10] designed an optimization model for joint routing, power allocation and spectrum assigning problem and illustrated the differences between routing in MCMRNs and that in CRAHNs. Filippini et al. in [11] proposed a minimum maintenance cost routing for cognitive radio networks. The authors formulated the maintenance cost problem to be an integer optimization model. By carefully selecting routing metric, the authors designed a heuristic distributing algorithm. All the above routing solutions rely heavily on the accuracy of the spectrum sensing algorithms at CR nodes. To ease this demand, in [12] Chowdhury et al. proposed a routing solution which avoided interference to PU receivers. In that solution, each time SUs chose the path which had minimum overlap with PUs' transmission areas. In [13] Ding et al. proposed a distributed algorithm to maximize the capacity of links without generating fatal interference to other users. In [14] Xie et al. proposed a greedy approach for relay selections which avoided causing harmful interference to PUs in CRAHNs. In [15] the authors Zhou et al. gave a mathematical model minimizing the interference to PUs and proposed an algorithm called Sequence Fixing to solve this problem. In [16] we set up a mixed integer linear programming (MILP) model to maximize SUs' throughput while keeping the interference caused to PU under a fixed threshold. In this paper, we mainly focus on finding a distributed algorithm to solve the MILP problem mentioned in [16]. To solve the problem, we design a proper routing metric which integrates both the interference to PU and the throughput of SUs. After selecting the path, we design the time-slot assignment algorithm to assign time-slots for the links in the path. The theoretical analysis shows that the average complexity of our proposed algorithm is O(T. 10gdN. log 10gdN), where T, d and N denote the time-slot period, the nodes' average degree and the number of nodes in the network respectively. The simulation results demonstrate that the proposed algorithm can approach the performance of centralized algorithms in SUs' throughput. The rest of this paper is organized as follows. In section 2, we present the system model and problem formulation. In section 3, we present the distributed algorithm and the complexity analysis. Section 4 demonstrates the simulation results and section 5 concludes this paper /12/$ IEEE

2 PU SUd SUe binary integral variables Xijt, where xijt=1 indicates that SU node i sends packets to SU node j in slot t; otherwise Xijt=O. Let fijt denote the actual flow rate in the link lij at time-slot t. Let Ai and Ii denote node i's neighbor nodes and interference nodes respectively. In such networks, our aim is to find the optimal routing and time-slot assignments which maximize the average throughput of SUs. Mathematically, we have the following optimization problem. SUb (1) Fig. 1. Network Architecture II. NETWORK ARCHITECTURE AND PROBLEM A. Network Architecture FORMULATION To improve the efficiency of TV frequency band, wireless devices are proposed by FCC to utilize television channel frequencies under the precondition of causing limited interference to PUs. Under this proposal, we consider a network consisting of 1 PU, N SUs. Each SU is equipped with two wireless transmitters: one called control transmitter works on a common control channel using special frequency bands, and the other called data transmitter shares TV frequency channel with PU. To avoid too much interference to PU, SUs' transmitting power is limited and so SUs are necessary to conduct multi-hop communications for long distance data transmitting. We use Fig. 1 to describe the network architecture. The TV receiver acts as a PU and has priority to use the spectrum f. Each active SUs broadcast its locations and its current using time-slot to its neighbors. The broadcast is conduct by control transmitter on CCC at regular intervals. Each time the SU node receives the broadcast packet it will update its available time-slots. The available time-slots are slots which SU itself can use while do not conflict with the other transmitting SUs. If the source node of SUs s wants to communicate with destination node SUs d, it first looks up its routing tables. If there is no route to destination it will send RREQ message on CCC by control transmitter. SUs want to find the path and time-slots which have high capacity while the cumulative interference dose not exceed PU's threshold. For exmple, in Fig.l source node s selects the route s ---t a ---t g ---t e ---t d. Node g is not selected because it will cause severe interference to PU. And then source node s will assign time-slot S ---t a ---t b ---t C ---t e ---t d for the find path. The symbol "a b" represents that node a transmits to node b in time slot 1. This paper focuses on finding a distributed routing and time-slot assignment which maximizes the SUs' throughput while avoiding harmful interference to PUs. B. Problem Formulation To describe the network mathematically, we first model the successful transmitting conditions among nodes and introduce s.t. (2) N 2: XijtgikQ::::; In T, j E Ai, k E P,O::::; t::::; T i=l Xijt = 0 or 1 fijt )! 0 Where B, T, Q, 7J and InT are all constants and denote frequency band width, time-slot periods, SUs' transmitting power, noise power and PU's threshold respectively. Note that the above optimization problem has both integral and continuous variables and the constraint conditions are all linear. So it is a kind of MILP problem. In the above MILP problem, the optimization variables consists of continuous variables fijt and binary variables Xijt. In [16] we have proposed a near optimal centralized solution RSAA to this problem. Here, we focus on the distributed algorithm. III. DRTSA ALGORITHM The purpose of this paper is to find a solution to maximizing SUs' throughput while keep SUs' cumulative interference to PU under a fixed threshold. The problem can be divided into two separate problems: path selection and time-slot assignment. The objective of path selection is to find a path from the source to the destination and the path should keep away from PU's location while providing high throughput for SUs. Time-slot assignment is designed to avoid concurrent collisions among SUs and reduce interference to PU. A. Path Selection Process We take two factors for each link into accountlink's interference to PU denoted as In(lij) and link's capacity denoted as Cap(lij). The path selecting algorithm is designed to select the path P which has high capacity and low interference to PU. So we define the cost of link lij as LCost(lij) = In(lij) l -a Cap(lij)'",0::::; a::::; 1 (3) where a is a tunable parameter. When a becomes 0, the metric will become the least overlap area with the PU as [12] depicted. And the metric proposed in [12] considers interference 471

3 to PU, but neglects SUs' throughput. When a becomes 1, the metric will become maximal cumulative capacity. The cost of path P is the sum of each link's cost and can be expressed as Steps TABLE I TIME-SLOT ASSIGNMENT ALGORITHM Contents PCost(P) = L LCost(lij) (4) l;jep When a source node wants to communicate with its destination node, it first looks up its route tables. If there is no route to the destination, it will conduct route finding process in our proposed algorithm DRTSA. The route finding process is alike that in AODV[17], but there are some modifications should be done in HELLO Package and RREP Package to accomplish the routing and time-slot assignment other than merely change the metric Hop Counts by LC ost in RREQ Package. 1) HELLO Package: In DRTSA, the active SUs broadcast HELLO Packages periodically to "tell" the other SUs its location and its current using time-slots on the CCC. After receiving the HELLO Package, each SU checks whether itself will conflict with the current active SU by the location information in HELLO Package. If the active node will conflict with the current node, the node will pick the time-slot which used by the active nodes out from its available time-slots. 2) RREP Package: The destination node always selects the path with the minimal cost as the path from source to destination. After the path is selected, destination node will send a RREP Package to the source node. Each intermediate node who forwards this RREP will add its own location and available time-slots to the RREP Package. As long as the source node receives the RREP, the path is set up. The route finding process's function is to find the potential route from source node to destination node. When the source node receive the RREP it still could not send data to destination node because the source node still do not know which time-slot it should access. Randomly selected time-slots will cause collisions among SUs and high interference to PU. So after receiving the RREP, the source node should run the timeslot assignment algorithm. B. Time-slot Assignment Algorithm After selecting the path P, the source node will know each intermediate node's location and available time-slots by the REP message. We only take channel's large scale fading mto account and use geographic location to calculate the link capacity. By Shannon's Theory we have Const _n _ Cap(lij) - B 10g( d(lij) ), lij E P (5) TJ where n is the channel fading factor and Canst is a constant determined by node's transmitting power, d(lij) denotes the transmitting distance of link lij. We define the average capacity for link lij as AvrCap(lij) and it can be expressed as 1 T AvrCap(lij) L = T f ijt i=l And the time-slot assignment algorithm is as following: (6) step 1 step 2 step 3 step 4 0, calculate each link's Initialize Vlij E P, AvrCap(lij) = capacity Cap(lij). Update AvrCap(lij), find the bottleneck link of path P, lij = arg min {AvrCap(lij)}. lijep If there is a time-slot meet that the summation of all the SUs' interference to PU plus In(lij) is smaller than InT, then let fijt Cap(lij), remove t from the available time-slot of the = links which conflict with lij, back to Step 2; else go to Step 4. The ca acity of path's bottleneck link could not be augmented, return bme-slot assignment results and the algorithm end. After the path selected and time-slot assigned, the routing from source to destination is actually set up and the source node can transmit data to destination node on data transmitter. During the transmitting, each active node should broadcast HELLO Package to its neighbors and claim that the slots have been occupied. C. Complexity Analysis Next we will analyze the complexity of the distributed routing and time-slot assignment (DRTSA) algorithm. As we all know that the original MILP problem is N-P Hard and the complexity of finding the optimal solution expands exponentially as the number of nodes increases. DRTSA algorithm consists of two separate parts. One is path selection and the other is time-slot assignment. Because our path selection algorithm is distributed and AODV [17]alike, the computing complexity of path selecting can be neglected. And the complexity of DRTSA mainly lies in the time-slot assignment algorithm. When we do time-slot assignment for path P, the worst case is that the path length is equal to the number of links 1. In Step 2, the algorithm finds the minimal value of N - AvrCap among all the links. And in Step 3, the algorithm has to select the proper time-slot among all the T slots for each link. So the total complexity of the time-slot algorithm in the worst case is O(NTlogN). In general, if we denote the nodes' aver ge degree as (1, the average path length L can be denoted as L = log a N and the whole complexity can be expressed as O(T. log a N. log log a N). IV. PERFORMANCE EVALUATION In this section, we present simulation results to evaluate our solutions. We compare the network throughput with centralized RSAA algorithm [16]and Sequence Fixing algorithm[l 5], and distributed Min-Hop algorithm and Minoverlapping algorithm[12]. We use CPLEX to solve Sequence Fixing problem and our simulation platform is Matlab. Because in the system architecture the access scheme is TDMA which is collision avoiding when the time-slots are properly assigned, so Matlab platform is enough to demonstrate the algorithms' performance. We next describe the simulation scenarios and then discuss simulation results. 472

4 A. Simulation Scenarios Our simulation scenario is set at the rural and mountainous areas where the TV broadcast spectrum is underutilized and the SUs can transparently use these spectrums without generating harmful interference to PUs. The simulation area is a ring region and the inner and outer radius are 300 meters and 500 meters respectively. The specific simulation parameters are in Table II. TABLE II SIMULATION PARAMETERS O O O O Node density (nodes/km 2 ) Simulation Parameters The topology area The distribution of SUs' location Channel Propagation model The transmitting range of SUs The interfering range of SUs The band width of PUs The power of noise The transmitting power of SUs a Simulation times Values 0.5 Km2 Ring Region Uniform Distribution Raleigh Fading model 100m 120m 8MHz -l1odbw lw Fig. 2. SUs' throughput vs node density B. Simulation Results Fig.2 shows the trend of SUs' throughput when the node's density increases in the network. In Fig.2, the time-slots period we assign is 6 and PU's threshold is dBW. We define SUs' throughput as the average capacity of the path's bottleneck link. From Fig.2 we can see that our proposed DRTSA can obtain the second largest throughput for SUs and is just lower than the centralized algorithm RSAA. The Sequence Fixing algorithm can get the least throughput because it does not take the actual background into consideration and just solve the math problems. The Min-Hop and Min-overlap algorithm can obtain nearly the same throughput because Min Hop does not take the interference to PU into consideration while Min-overlap neglects the link capacity. Fig.3 shows the trend of SUs' throughput as PU's threshold increases. In Fig.3, the node's density is 44 nodes per Square Kilometers and the time-slot is 4. It can be seen that when PU's threshold is low all the five algorithms have nearly the same performance, but when PU's threshold increases the algorithms' performance varies. From Fig.2 and Fig.3 we can conclude that DRSTA algorithm's performance can approach that of the centralized algorithm in various scenarios. V. CONCLUSION In this paper, we proposed a distributed routing and timeslot assignment algorithm that addresses the problems of PU protection. Firstly, we design a proper metric for path selection and then we propose a time-slot assignment algorithm both takes SUs' capacity and interference to PU into account. Our performance evaluation shows that the proposed algorithm DRTSA can approach the centralized algorithm's performance whenever the nodes' density or PU's threshold varies. So the algorithm can be wildly used in cognitive radio ad hoc networks. O -L -L L--L o PU threshold (le-10 W) Fig. 3. SUs' throughput vs PU' s threshold ACKNOWLEDGMENT This work was supported in part by the National Natural Science Foundation of China ( ), and the National Major Special Projects in Science and Technology of China (2011ZX ), and the National High-Tech Development Program (2011AAOlA105). REFERENCES [1] J. Mitra III and G. Q. Maguire JR., "Cognitive radio: making software radios more personal," IEEE Personal Commun., pp , Aug [2] I. Akyildiz, W. Lee, M. Vuran, and S. Mohanty. "NeXt Generation / Dynamic Spectrum Access/Cognitive Radio Wireless Networks: A Survey." Compo Netw. Jour. (Elsevier), Vol.50, no. 13, pp , Sept [3] I. Akyildiz, W. Lee, and K. Chowdhury. " CRAHNs: Cognitive Radio Ad Hoc Networks" Ad Hoc Networks Journal (Elsevier), Vol.7, no.5, pp , July [4] T. Yucek, H. Arslan, "A survey of spectrum sensing algorithms for cognitive radio applications," Communications Surveys & Tutorials, IEEE, vol. 11, no.l, pp , First Quarter [5] H. Wang, H. Qin, L. Zhu, "A Survey on MAC Protocols for Opportunistic Spectrum Access in Cognitive Radio Networks," Computer Science and Software Engineering, 2008 International Conference on, vou, no., pp , Dec [6] Y. Wang, P. Ren, and G. Wu, "A throughput-aimed MAC protocol with QoS provision for cognitive ad hoc networks," IEICE Trans. Commun., vol. E93-B, no. 6, pp , Jun [7] H. Khalife, N. Malouch, S. Fdida, "Multihop cognitive radio networks: to route or not to route," Network, IEEE, vol.23, no.4, pp.20-25, July August 2009 [8] X. Zhou, L. Lin, J. Wang and X. Zhang, "Cross-layer routing design in cognitive radio networks by colored multigraph model;' Wireless Personal Communications, vol. 49, no.l, pp , April

5 [9] Y.T. Hou, Y. Shi, H.D. Sherali, "Optimal Spectrum Sharing for Multi Hop Software Defined Radio Networks," INFOCOM th IEEE International Conference on Computer Communications. IEEE, vol., no., pp.i-9, 6-12, May 2007 [10] Y.T. Hou, Y. Shi, H.D. Sherali, "Spectrum Sharing for Multi-Hop Networking with Cognitive Radios," Selected Areas in Communications, IEEE Journal on, vo1.26, no.l, pp , Jan [11] I. Filippini, E. Ekici, M. Cesana, "Minimum maintenance cost routing in Cognitive Radio Networks," Mobile Adhoc and Sensor Systems, MASS '09. IEEE 6th International Conference on, vol., no., pp , Oct [12] K.R. Chowdhury, I.E Akyildiz, "CRP: A Routing Protocol for Cognitive Radio Ad Hoc Networks," Selected Areas in Communications, IEEE Journal on, vol.29, no.4, pp , April 2011 [13] L. Ding, T. Melodia, S.N. Batalama, J.D. Matyjas, M.J. Medley, "Cross Layer Routing and Dynamic Spectrum Allocation in Cognitive Radio Ad Hoc Networks," Vehicular Technology, IEEE Transactions on, vol.59, no.4, pp , May 2010 [14] M. Xie, W. Zhang, K.K. Wong, "A geometric approach to improve spectrum efficiency for cognitive relay networks;' Wireless Communications, IEEE Transactions on, vol.9, no.l, pp , January 2010 [15] Z. Yuan, J. B. Song, Z. Han, "Interference Minimization Routing and Scheduling in Cognitive Radio Wireless Mesh Networks," Wireless Communications and Networking Conference (WCNC), 2010 IEEE, vol., no., pp.i-6, April 2010 [16] H. Chen, Q. Du, P. Ren, "A Joint Routing and Time-Slot Assignment Algorithm for Multi-Hop Cognitive Radio Networks with Primary-User Protection" International Journal on Computers, Communications & Control, Vol.7,No.3,pp , sept [17] C.E. Perkins and E.M.Royer, "Ad hoc on-demand distance vector routing". In Proc. IEEE Workshop on Mobile Computing Systems and Applications (WMCSA99), New Orleans, USA, 1999: